Operator-Space Mutual Information (OSMI)

• OSMI is a measure from free probability theory that quantifies operator spreading in quantum systems by aggregating all higher moments of correlations.
• It refines chaos diagnostics by integrating the full hierarchy of even-point OTOCs to assess ergodicity and quantum scrambling.
• OSMI finds application in both quantum chaos analysis and random matrix models to differentiate true scrambling from integrable dynamics.

Operator-Space Mutual Information (OSMI), also referred to as Free Mutual Information (FMI) in the physics literature, is a rigorous quantity adapted from free probability theory that quantifies the extent of operator spreading in quantum systems. OSMI refines standard mutual information and chaos diagnostics by taking into account all higher moments of operator correlations, making it a key figure of merit for characterizing ergodicity and quantum information scrambling in large Hilbert spaces and random matrix ensembles (Vardhan et al., 16 Sep 2025, Diaz et al., 2014).

1. Definition and Mathematical Formulation

OSMI generalizes the classical mutual information of random matrices to the operator-valued setting and is designed to probe operator growth and delocalization in the space of all linear operators on the Hilbert space. Let $A(t)$ be the Heisenberg-evolved operator and $B$ a reference observable. The space of unitary rotations of $A$, denoted $\mathcal O_A = \{UAU^\dagger | U \in U(d)\}$, is exponentially large in the Hilbert space dimension $d$.

The Free Mutual Information $I_F(A(t):B)$ is defined as

$I_F(A(t):B) \equiv -\frac{4}{d^2} \log f_{A(t)|B}$

where $f_{A(t)|B}$ is the volume fraction of operators in $\mathcal O_A$ matching all joint moments with $B$ up to some truncation. Equivalently,

$$I_F = \frac{4}{d^2}\left[\log\operatorname{Vol}(\mathcal O_A) - \log\operatorname{Vol}(\mathcal O_{A(t)|B})\right]$$

For involutive operators ($A,B$ with spectrum $\pm1$), an explicit Coulomb-gas formula in terms of the eigenvalues $z_j$ of $A(t)B$ holds:

$$I_F(A(t):B) = -\frac{4}{d^2}\sum_{1 \leq i < j \leq d/2} \log|x_i - x_j| - \log 2 + o(1)$$

where $x_j = \mathrm{Re}\,z_j = \cos\theta_j$ (Vardhan et al., 16 Sep 2025).

In operator-valued random matrix models, given $\mathbf H$ in $M_{n_R}(\mathcal C)$ in a noncommutative $C^*$-probability space, the asymptotic operator-space mutual information is

$$I_{OSMI} = \int_0^\infty \log(1 + P\xi)\,dF_{\mathbf H\mathbf H^*}(\xi)$$

where $F_{\mathbf H\mathbf H^*}$ is the analytic distribution of the operator $\mathbf H\mathbf H^*$ (Diaz et al., 2014).

2. Connections to Out-of-Time-Ordered Correlators (OTOCs)

OSMI is intimately connected to the hierarchy of out-of-time-ordered correlators, generalizing diagnostics of quantum chaos beyond the conventional four-point OTOC. The key relation is the decomposition

$$I_F(A(t):B) \simeq \sum_{n=1}^\infty \frac{2}{n}\, \mathrm{OTOC}_{2n}(t)^2$$

where $$\mathrm{OTOC}_{2n}(t) = \frac{1}{d}\operatorname{Tr}[(A(t)B)^n]$$ (Vardhan et al., 16 Sep 2025). This formula accumulates information from all even-point correlators, providing a measure of operator ergodicity and the approach to freeness. Oscillations in individual higher-point OTOCs are smoothed out in the sum, ensuring that $I_F$ robustly signals operator-space delocalization.

In classical MIMO random matrix channels, OSMI similarly aggregates higher moments of the block-structured channel matrix, extending mutual information estimates to encompass noncommutative operator-valued distributions (Diaz et al., 2014).

3. Operator-Valued Kronecker Models and Analytic Techniques

To analyze OSMI in random matrix systems with correlated blocks (e.g., antenna pattern correlations in MIMO models), the operator-valued Kronecker model is introduced:

$\mathbf H = \mathbf R\,\mathbf X\,\mathbf T$

with diagonal block correlation operators $\mathbf R$, $\mathbf T$, and a matrix $\mathbf X$ of *-free circular elements. The distribution of $\mathbf H\mathbf H^*$ is characterized using operator-valued Cauchy transforms and free-probability subordination methods. The symmetrized operator

$$\widehat{\mathbf H} = \begin{pmatrix}0 & \mathbf H\ \mathbf H^* & 0\end{pmatrix}$$

encodes the spectrum of interest, and analytic subordination yields fixed-point equations for computing the mutual information via Stieltjes inversion (Diaz et al., 2014).

In the fully diagonal case, the machinery reduces to solving $2n$ scalar equations iteratively:

• Compute $Q_k = G_{r_k^2}(d_k)$ or $G_{t_{k-n}^2}(d_k)$,
• $$x_k = \frac{1}{d_k - \frac{1}{2n}\sum_{\ell=1}^{2n}\Sigma^2_{k\ell} x_\ell}$$,
• Fixpoints $\omega_{2,k}$ via $g_k$ maps and associated $h$-transforms,
• Recover the scalar Cauchy transform and integrate for $I_{OSMI}$.

This direct algorithm is numerically validated in block-structured random matrix simulations.

4. Physical and Information-Theoretic Interpretation

OSMI quantifies ergodicity and scrambling in operator space. As $A(t)$ evolves, its support explores an exponentially large operator manifold $\mathcal O_A$, and $I_F$ tracks the fraction of this space excluded by residual correlations with $B$. In chaotic quantum systems, this volume grows until it fills $\mathcal O_A$, signaling maximal delocalization and noncommutativity (freeness). In contrast, localized or integrable regimes yield $I_F \to \infty$, indicating persistent sub-algebra structure and restricted operator spreading.

Physically, OSMI serves as the "mutual information across time", but in the noncommutative (free probability) framework rather than classical probability. It refines operator growth diagnostics—whereas standard OTOC measures spatial extent, OSMI diagnoses the ultimate scrambling in the vast operator manifold, which is doubly exponential in system size.

5. Universal Behaviors in Quantum Chaos and Random Matrix Ensembles

Comprehensive studies reveal universal OSMI behavior across a range of chaotic many-body models:

• Random GUE Hamiltonians: $I_F(t) > 0$ for any $t > 0$; rapid decay to $O(1/d^\alpha)$ on short timescales; higher-point OTOCs display fast initial decay.
• Brickwork Haar-random unitary circuits: Lieb–Robinson lightcone structure, with all OTOC$_{2n}$ decaying simultaneously at the front velocity; $I_F(t,x)$ follows a sharp transition.
• Local chaotic spin chains: After ballistic front propagation, OSMI matches random-circuit predictions for operator spreading.
• Integrable and pseudorandom systems: Clifford/PFC ensembles and certain free-fermion models exhibit $I_F = \infty$ regardless of operator evolution, despite finite four-point OTOC decay.
• Operator-valued Kronecker models: Empirical simulations of large random block-matrix channels corroborate analytic OSMI predictions; realistic antenna correlation models demonstrate deviations from classical Kronecker behavior (Diaz et al., 2014).

Model Class	OSMI Behavior	Notable Feature
Chaotic (GUE, spin chain)	$I_F$ decays	Doubly-exponential operator-space spreading
Integrable, pseudorandom	$I_F = \infty$	No operator-space delocalization
Block-correlated random MIMO	OSMI sensitive to fine-scale correlation	Explains classical deviations

6. Applications and Experimental Accessibility

OSMI is a sharper diagnostic of quantum chaos and information scrambling than conventional measures. It enables direct quantification of operator-space ergodicity, distinguishing genuine chaos from pseudorandom delocalization. The measure is experimentally accessible: higher-point OTOC$_{2n}$ (with $n \leq 4$) have already been measured in random-circuit quantum computing platforms; OSMI can be reconstructed from these correlators (Vardhan et al., 16 Sep 2025).

In wireless information theory, OSMI provides asymptotically exact estimates of channel capacity under complex block-based and angle-diversity models, outperforming scalar approaches especially when strong pattern correlation is present (Diaz et al., 2014).

7. Comparison, Limitations, and Outlook

OSMI unifies perspectives from quantum chaos (free probability, operator ergodicity) and random matrix theory (asymptotic mutual information in block channels). It refines operator growth diagnostics by condensing the entire hierarchy of even-point OTOCs into a single ansatz-free indicator. OSMI distinguishes true operator-space scrambling from mere spatial delocalization (where $I_F$ may diverge despite OTOC decay), making it invaluable for characterizing quantum dynamics beyond integrability and pseudorandomness.

A plausible implication is that, for comprehensive characterization of scrambling and entropy growth in quantum many-body systems or complex communication channels, monitoring OSMI and the underlying higher-point correlator structure is indispensable. The robustness of OSMI in both numerical simulation and experimental contexts, and its analytic tractability via free-probability methods, position it as a central tool in modern quantum information theory and applied random matrix research (Vardhan et al., 16 Sep 2025, Diaz et al., 2014).